THE ANATOMICAL RECORD 294:1472–1485 (2011)

Analysis of the Effect of Osteon Diameteron the Potential Relationship of

Osteocyte Lacuna Density and OsteonWall Thickness

JOHN G. SKEDROS,1* GUNNAR C. CLARK,1 SCOTT M. SORENSON,1

KEVIN W. TAYLOR,1 AND SHIJING QIU2

1Department of Veterans Affairs Medical Center, Bone and Joint Research Laboratory,and University of Utah, Salt Lake City, Utah

2Bone and Mineral Research Laboratory, Henry Ford Hospital, Detroit, Michigan

ABSTRACTAn important hypothesis is that the degree of infilling of secondary

osteons (Haversian systems) is controlled by the inhibitory effect of osteo-cytes on osteoblasts, which might be mediated by sclerostin (a glycopro-tein produced by osteocytes). Consequently, this inhibition could beproportional to cell number: relatively greater repression is exerted byprogressively greater osteocyte density (increased osteocytes correlatewith thinner osteon walls). This hypothesis has been examined, but onlyweakly supported, in sheep ulnae. We looked for this inverse relationshipbetween osteon wall thickness (On.W.Th) and osteocyte lacuna density(Ot.Lc.N/B.Ar) in small and large osteons in human ribs, calcanei ofsheep, deer, elk, and horses, and radii and third metacarpals of horses.Analyses involved: (1) all osteons, (2) smaller osteons, either �150 lm di-ameter or less than or equal to the mean diameter, and (3) larger osteons(>mean diameter). Significant, but weak, correlations between Ot.Lc.N/B.Ar and On.W.Th/On.Dm (On.Dm ¼ osteon diameter) were found whenconsidering all osteons in limb bones (r values �0.16 to �0.40, P < 0.01;resembling previous results in sheep ulnae: r ¼ �0.39, P < 0.0001). Inlarger osteons, these relationships were either not significant (five/sevenbone types) or very weak (two/seven bone types). In ribs, a negative rela-tionship was only found in smaller osteons (r ¼ �0.228, P < 0.01); thisinverse relationship in smaller osteons did not occur in elk calcanei.These results do not provide clear or consistent support for the hypothe-sized inverse relationship. However, correlation analyses may fail todetect osteocyte-based repression of infilling if the signal is spatially non-uniform (e.g., increased near the central canal). Anat Rec, 294:1472–1485, 2011. VVC 2011 Wiley-Liss, Inc.

Keywords: osteons; osteocytes; osteon wall thickness; boneremodeling; sclerostin

Grant sponsor: NIH; Grant number: DK43858; Grantsponsor: Department of Veterans Affairs; Grant sponsor:Orthopaedic Research and Education Foundation (OREF;medical research fund); Grant number: 01-024; Grant sponsor:Utah Bone and Joint Center, Salt Lake City, UT.

*Correspondence to: John G. Skedros, M.D., Utah Bone andJoint Center, 5323 South Woodrow Street, Ste. 202, Salt Lake

City, UT 84107. Fax: þ1-801-713-0609. E-mail: jskedros@utah-boneandjoint.com

Received 1 February 2011; Accepted 20 June 2011

DOI 10.1002/ar.21452Published online 1 August 2011 in Wiley Online Library(wileyonlinelibrary.com).

VVC 2011 WILEY-LISS, INC.

The process of remodeling in cortical and cancellousbone is important for ensuring the maintenance of me-chanical competence and calcium balance (Frost, 1990;Parfitt, 1994). In this context, ‘‘remodeling’’ refers to theformation of complete secondary osteons (Haversian sys-tems) in cortical bone and hemiosteons in cancellousbone (Parfitt, 1994). The remodeling process is charac-terized by the coupled actions of osteoclasts (boneresorption) and osteoblasts (bone formation). These cellscomprise the basic multicellular units that mediate theactivation–resorption–formation (A–R–F) sequence thatresults in secondary osteons, or basic structural units(Frost, 2001; Cooper et al., 2006). Osteocytes, the mostprevalent cell type in bone, are thought to be importantin mechano-sensation/mechano-transduction (Marottiet al., 1992; Burger and Klein-Nulend, 1999; Weinbaumet al., 2001; Riddle and Donahue, 2009; Herman et al.,2010).

Although most osteoblasts involved in osteon forma-tion are not accounted for at the termination of infilling(they presumably undergo apoptosis; Parfitt, 1994), theremainder differentiate into osteocytes as they becomeembedded in the newly formed matrix; their cell bodiesbeing contained within osteocytic lacunae (ellipsoid cav-ities). Marotti (1996) hypothesized that during the for-mation of secondary osteons, the newly formedosteocytes send inhibitory signals to surrounding osteo-blasts. This signal causes the osteoblasts to cease boneformation. Recent studies of the mechanical stimulationof bone have suggested that this inhibitory effect is regu-lated by the expression of the SOST gene, which is re-sponsible for the formation of sclerostin, an osteocyte-specific glycoprotein involved in the inhibition of osteo-blasts (Van Bezooijen et al., 2004; Robling et al., 2008;Turner et al., 2009). Experimental evidence suggeststhat sclerostin is produced by osteocytes that are withinwell mineralized matrix (Irie et al., 2008).

Before recognition of the importance of sclerostin inbone remodeling, Martin expanded on this osteocyte in-hibitory concept with his ‘‘unifying theory of boneremodeling,’’ wherein osteoblast activity is controlled byosteocytes via signaling through their interconnectednetwork. More specifically, osteoblasts remaining at thetermination of osteon infilling (i.e., termination of osteonwall formation) differentiate into a distinct cell type—bone lining cells (BLCs; Martin, 2000a,b). In turn, theBLCs maintain communication with the embeddedosteocytes via gap junctions (Fig. 1). At the terminationof osteon infilling, instead of regulating osteocyte forma-tion, the interconnected osteocytes, in theory, send inhib-itory signals to the BLCs to suppress the activation ofthat would lead to a new remodeling cycle (Stages 8 and1 in Table 1). Martin (2000b) further hypothesized thatthe strength of the inhibitory signal is proportional tothe number of osteocytes in communication with theBLCs.

Experimental data reported by Metz et al. (2003) inulnae of adult sheep provide indirect support for theidea that the osteocyte network regulates the rate andduration of osteon infilling (Fig. 2). However, althoughtheir data revealed an inverse relationship betweenosteon wall thickness (On.W.Th) and osteocyte lacunadensity (Ot.Lc.N/B.Ar; B.Ar ¼ bone area within a sec-ondary osteon), it was weak (r ¼ �0.39, P < 0.0001).Nevertheless, these authors concluded that it was biolog-

ically important, supporting the hypothesis that osteo-cytes influence the degree of osteon infilling. AlthoughQiu et al. (2003) also argued in favor of this relationshipin their analysis of osteons from ribs of young humanmales, they did not specifically consider On.W.Th, butonly evaluated relationships of Ot.Lc.N/B.Ar withHaversian canal area (HC.Ar) and HC.Ar/On.Ar (On.Ar¼ osteon area). Both of these comparisons in human ribosteons, however, showed no or little correlation (r ¼0.011 and 0.104, respectively), and only the latter is pos-itive and statistically different from no correlation.Regardless of the specific mechanism(s) involved, Martinand coworkers’ important hypothesis has not been exam-ined in other bones and/or for potential influences ofosteon size. The only studies that we could locate haveexamined osteons in cortices of elderly human proximalfemora (no consideration for osteon size; Power et al.,2001, 2002), which could be confounded by the effects ofbone disease (e.g., osteoporosis) and/or aging (Britzet al., 2009).

Influences of osteon size on the putative inverse rela-tionship of Ot.Lc.N/B.Ar with On.W.Th could occur inproportion to the efficiency with which the proposed in-hibitory signal can be transmitted from one osteocyte toanother (e.g., through a neural-like network) or deliv-ered (e.g., convective transport of sclerostin in lacunar–canalicular spaces) across the osteon wall. In a mathe-matical analysis dealing with the duration and extent ofosteon infilling, Martin (2000a) argued that if the osteo-cyte-network signal decayed slowly as the osteon pro-gressively infilled, then the strength of the inhibitorysignal at the canal surface would depend on osteon size.But if the signal decayed relatively rapidly, then the sig-nal would not increase very much once an osteon’s wallexceeds �100–120 lm (Fig. 3), constraining potential

Fig. 1. Schematic diagram of the proposed mechanism for the con-trol of activation of bone remodeling. Within the bone tissue at ‘‘top,’’the elliptical objects represent osteocytes. The heavy line at ‘‘bottom’’indicates a quiescent (not remodeling) bone surface; five BLCs areshown on this surface. The multiple thin lines connecting the osteo-cytes to one another and to the BLCs schematically represent cellprocesses within canaliculi. The jagged horizontal line indicates micro-damage (microcrack) to the calcified matrix, which interferes with thegeneration or transmission of the osteocytic signal. Reduced signalgeneration/transmission is indicated as dotted lines for the canaliculiphysically disrupted by the microcrack. In theory, when these inhibi-tory signals fall below a threshold value, the BLCs initiate activation ofremodeling. (From Martin, Bone, 2000b, 26, 1–6, Fig. 2, VC Elsevier,reproduced by permission.)

OSTEOCYTE DENSITY AND OSTEON WALL THICKNESS 1473

increases in osteon size. Similar arguments could bemade on the basis of the ‘‘transport’’ of sclerostin viaconvective fluid flow through the interstitial microporosi-ties of the osteon wall.

Regardless of the nature of the putative inhibitory‘‘signal’’ and how it is ‘‘transmitted’’ or ‘‘delivered,’’ wehypothesize that if it is produced by osteocytes and isnegatively correlated with the degree of osteon infilling,then in fully formed osteons: (1) it should be detected asan inverse relationship between On.W.Th and Ot.Lc.N/B.Ar and (2) this inverse relationship should be influ-enced by osteon size.

MATERIALS AND METHODS

As detailed below, microscopic images were obtainedfrom the ribs of young adult modern human males (20–25 years old) and various nonhuman limb bones fromadult animals: (1) radii and third metacarpals (MC3s)from horses and (2) calcanei from horses, sheep, deer,and elk. These nonhuman bones had been used in ourprevious studies (Skedros et al., 1994a,b, 1996, 1997,2007b; Mason et al., 1995; Skedros, 2005). The humanribs had also been used in a previous study (Qiu et al.,2003). The rationale for selecting the various nonhumanbone types is that they reflect three different loadingregimens that correlate with the magnitude of neutralaxis rotation during typical functional loading (Fig. 4).In turn, these regimens represent differences in loadcomplexity from the simple cantilevered bending of the

calcanei to the increased bending/torsion of the equineMC3s. The rationale for examining the ribs is based onthe fact that a great deal of what is known about basicosteon dynamics has been determined in the studies ofribs (Pirok et al., 1966; Parfitt, 1983; Qiu et al., 2003).Finally, the osteons in these bones also span a broad sizerange, which, as argued in our previous studies andthose of others (Pfeiffer et al., 2006), reflects adaptationsaimed at enhancing tissue mechanical properties for re-gional differences in mechanical requirements (limbbones) and/or metabolic demands (ribs).

The regions examined in the nonhuman bones areshown in Figs. 5 and 6. It must be noted that below weprimarily use ‘‘anterior’’ and ‘‘posterior,’’ respectively, assurrogates for the actual, but various, terms used for theopposing sagittal-plane cortices in the calcanei (dorsaland plantar), radii (cranial and caudal), and equine MC3(dorsal and palmar).

Equine Radii and MC3s

Ten radii and nine third metacarpals (MC3s) wereobtained from skeletally mature standard-bred andquarter horses with no history of racing or race trainingand no evidence of skeletal pathology at the time ofdeath (Figs. 4 and 6). Only one bone was obtained fromeach animal. The bones had been used in previous stud-ies reporting correlations of various material character-istics with habitual strain environments from the mid-third of the bone (radius) or at mid-diaphysis (MC3;

TABLE 1. Stages of the remodeling process with the descriptions given for the most salientcharacteristic(s) of each stage

Stages of remodeling (the formation of secondary osteons)

Activation Resorption Formation

1 2 3 4a

Activation Migration & proliferation Bone resorption Bone formation

The beginning of theprocess of remodeling.

Osteoclasts are recruited,migrate, and proliferate.

Formation of aresorption cavityby osteoclasts.

Osteoblasts form boneon the resorbed surfaceand lay down sequentiallamellae. The cement lineis formed first.

Formation Quiescent stageb

5a 6a 7 8Differentiation Termination Bone lining cells Possible repressionOsteoblasts differentiateinto osteocytes,which become embeddedin the bone matrix.

Formation of the central canaldue to termination of infilling.(Strongly influenced byosteoblast apoptosis.)

Blcs cover the thebone surface ofthe central canal.

Osteocytes, via control of BLC,repress activation of a newremodeling cycle (Fig. 1).

Stages 4–6 (the formation stage) are the focus of this study because they deal with the hypothesis that osteocytes densityis inversely related to the degree of osteon infilling. Another hypothesis is that osteocytes send a repressive signal to theBLCs that inhibits bone resorption (Stages 8 and 1; Fig. 1). This hypothesis was not evaluated in this study. The delinea-tion of eight stages, as done here, serves to clarify the focus of the data analysis in this study. They are not intended tosupplant the widely accepted quiescence period, then A-R-F sequence-each stage with its own well-understood and inte-grated cellular cascades.BLC: bone lining cells; central canal: Haversian canal; secondary osteon: Haversian system.aStages of remodeling addressed in the current study.bThe quiescent stage is defined by the remaining osteoblasts becoming lining cells, and then remaining as such for sometime. This stage is a complex integrative physiological state that involves osteocyte production of signals, BLC integrationof those signals, and then BLC communication to endothelial cells and osteoblast and osteoclast precursors.

1474 SKEDROS ET AL.

Mason et al., 1995; Skedros et al., 1996, 2005a). Eachbone had been manually cleaned of soft tissue and sec-tioned transversely at 50% of its length (MC3), or at50% and 65% of its length (radius), where 100% is theproximal end of the bone. One 5-mm-thick segment wascut immediately distal to these transection levels. Theundecalcified, unstained segments were embedded inpolymethyl methacrylate using conventional techniques(Emmanual et al., 1987), and they were subsequentlyground, polished, and prepared for backscattered elec-tron (BSE) imaging (Skedros et al., 1993).

In the MC3s, two-to-three 50� high-resolution BSEmicrographs (2.3 � 1.6 mm2) were taken within each ofthe three trans-cortical locations [i.e., periosteal(‘‘outer’’), middle cortical, endosteal cortex (‘‘inner’’)] inquadrant locations: anterior, posterior, medial, and lat-eral. In the radii, two 50� (2.71 � 2.71 mm2) BSE micro-graphs were taken within each of the three trans-cortical locations of the anterior ‘‘tension’’ and posterior‘‘compression’’ cortices, and one image in each trans-cort-ical location of the medial and lateral (‘‘neutral axis’’)cortices. The images excluded circumferential lamellarbone.

Calcanei of Horses, Elk, Sheep, and Deer

One calcaneus was obtained from each of the sevenstandard-bred horses with no history of racing or racetraining, seven wild-shot North American elk (Cervuselaphus), and seven domesticated sheep (Ovis aries,breed is crossed Suffolk/Hampshire and Rambouillet;Fig. 5). All of these bones were from adult animals thathad been used in previous studies (Skedros et al., 1997,2007b). All elk were larger males, had shed the perios-teal cover of their antlers, and were obtained from theirnatural habitat in a fall hunting season. The horse cal-canei were from mixed sexes and were not from thesame animals from which the radii and MC3s wereobtained. All horses were between 2 and 9 years old.

The sheep were females and �2 years old. The sheepand horses had been kept in large pastures. Calcaneiwere also obtained from 10 wild-shot skeletally maturemale deer used in the previous studies (Skedros et al.,1994a,b, 2007b; Skedros, 2005).

Transverse segments, 4–5-mm thick, were cut fromthe sheep, elk, and horse calcanei at 60% of shaft length.The deer calcanei were sectioned at both the 50% and70% locations. These unstained, undecalcified segmentswere embedded in polymethyl methacrylate using con-ventional methods (Emmanual et al., 1987). The proxi-mal face of each segment was ground, polished, andprepared for BSE imaging using the same techniquesused for the equine radii and MC3s. One 50� BSEimage representing 3.41 mm2 (2.23 � 1.53 mm2) wasobtained in each trans-cortical location (periosteal, mid-dle cortical, and endosteal) of the dorsal ‘‘compression’’(anterior) and plantar ‘‘tension’’ (posterior) cortices ofeach specimen. Because of narrow plantar cortices of thesheep calcanei, only two images were obtained there.One image was also taken within each medial and lat-eral cortex (‘‘neutral axis’’) of the sheep and horse calca-nei. The images excluded circumferential lamellar bone.

Histomorphometric Analyses of Human Ribs

The human osteon data were obtained from archivedhuman rib specimens that had been used by Qiu et al.(2003). These specimens had been obtained by Frost inthe early 1960s from the middle portions of the fifth,sixth, or seventh ribs of nine previously healthy 20–25-year-old white male cadavers. The method of specimenprocessing has been previously described (Frost, 1960).In brief, 3-in. (7.6 cm) rib segments were placed in 1%basic fuchsin and 40% ethyl alcohol for 4 weeks andthen immersed in a large volume of tap water for 48 hr.

Fig. 3. The hypothesized areal inhibitory signal (normalized by volu-metric signal strength), within a completed osteon, is plotted as afunction of the cement line radius (RC) for several values of the decayconstant k (mm-1). The radius of the HC is assumed to be 20 lm.When k surpasses about 25 mm-1, the inhibitory signal on the HC pla-teaus as the cement line radius (RC) exceeds 100 lm (0.10 mm on theabscissa). SNA is the inhibitory signal (SN) per unit area (subscript A)on the surface of the HC. (From Martin, Bone, 2000a, 26, 71–78, Fig.6, VC Elsevier, reproduced by permission.)

Fig. 2. Schematic representation of osteons that show the generalhypothesis supported by data from Metz et al. (2003) in mature sheepulnae. HC.Ar ¼ Haversian canal area (¼central canal area).

OSTEOCYTE DENSITY AND OSTEON WALL THICKNESS 1475

After the samples were hydrated, transverse sectionswere cut from each rib, ground to a thickness of about50 lm, and mounted on slides.

Microscopic analysis of the rib osteons and their osteo-cyte lacunae (Ot.Lc.N) was performed on the imagesobtained from the transverse cross-sections using aNikon microscope with a CCD video camera (Optronics,Goleta, CA). Each microscopic image was imported to aBioquant NOVA image analysis system (R&M Biomet-rics, Nashville, TN) with a panel sized 640 � 480 pixels.The length and width of the panel under 1� objectivewas 6.717 � 5.908 mm2. All of the secondary osteonsthat met the following criteria were examined: (1) anintact osteon with a clear boundary at the cement line,(2) a HC.Ar less than one-quarter of the On.Ar, with noosteoid on its surface, (3) no Volkmann’s canals crossingthe osteon and no aberrant or atypical osteon variant(e.g., unusually elliptical osteon, drifting osteon, anddumbbell osteon), (4) no interference from the basicfuchsin intensity with the measurement of Ot.Lc.N, and(5) staining intensity consistent with complete minerali-zation, in addition to the absence of osteoid. It was im-portant to select only mature osteons for this studybecause sclerostin is only produced by osteocytes thatare within well-mineralized matrix (Irie et al., 2008). Onthe basis of these criteria, a total of 398 osteons from

the nine human subjects were measured. In the ribs, theboundaries of the osteon and the Haversian canal (HC)were traced under bright light using a 10� objective,and then the stained lacunar profiles were countedunder blue-violet epifluorescent light using a 20� objec-tive. Ot.Lc.N can be readily identified under blue-violetlight due to the strong fluorescence of basic fuchsin.Because epifluorescent light penetrates <5 lm below thespecimen’s surface, a thin section can be assumed forcounting the Ot.Lc.N even when a section is thick(�100–150 lm; Mori et al., 1997; Vashishth et al., 2000;Slyfield et al., 2009). The image analysis system wasused to directly measure nearly all the osteon parame-ters [On.Ar, HC.Ar, and mineralized bone area of anindividual osteons (B.Ar)] from the digitized images.Using these measurements and assuming circularity ofthe osteons, mean On.W.Th was calculated for eachosteon as the difference between the outer (cement line)and HC radii (Fig. 7).

Fig. 5. All calcanei in medial-to-lateral view; from top to bottom:sheep, deer, elk, and horse. Cross-sections show image locations ingray. One section (60% location) was analyzed in sheep, elk, andhorse bones, and two sections (50% and 70% locations) were ana-lyzed in deer bones. Dor: dorsal; Pla: plantar; Med: medial; and Lat:lateral. The inset drawing at the upper right shows an adult right deercalcaneus in lateral-to-medial view showing typical loading along theAchilles tendon (large arrow) producing net compression (‘‘C’’) andtension (‘‘T’’) on dorsal and plantar cortices, respectively. This loadingregime is similar in all of the calcanei used in this study. In this figure,the ‘‘Achilles tendon’’ represents the common calcaneal tendon.

Fig. 4. Lateral-to-medial views of the right forelimb and hind-limbskeletons of an adult horse showing a spectrum from simple loading tocomplex loading, respectively: calcaneus (A), radius (B), and third meta-carpal (MC3; C). The drawings below are simplified renditions of eachbone type, showing (A) the calcaneus as a cantilevered beam and (B) theradius as a curved beam with longitudinal loading; the curvature accen-tuates bending. Torsion (dotted line) is also present but is less than thetorsion in the MC3 (solid circular line in C), and (C) the MC3 with off-axislongitudinal loading producing bending and torsion, the latter beinggreater than in the other two bones. Several studies reporting in vivostrain data were used to create these drawings (Lanyon, 1974; Turneret al., 1975; Rubin and Lanyon, 1982; Schneider et al., 1982; Bieweneret al., 1983a,b; Gross et al., 1992; Su et al., 1999).

1476 SKEDROS ET AL.

Histomorphometric Analyses of NonhumanBones

We also evaluated complete and mature secondaryosteons that had been identified in the BSE images fromthe transverse sections of the nonhuman limb bones inour previous studies. Maturity was determined in theseimages by gray levels of the osteonal walls that are in-dicative of complete mineralization (e.g., not relativelydarker as seen in forming, or recently formed, osteons;Skedros et al., 2007a). Each nonhuman osteon was

selected using the same criteria described above for therib osteons, except for criteria #4 and #5 that involvedbasic fuchsin because no staining had been done on thenonhuman specimens. The Ot.Lc.N were manuallycounted in each osteon. Ot.Lc.N can be readily identifiedin the BSE images due to the stark contrasts betweenthe black lacunae and gray bone. Furthermore, becausethe electrons that form the BSE images used in thisstudy penetrate only a few microns below the specimen’ssurface (Skedros et al., 2005b), a thin section can beassumed for measuring Ot.Lc.N even when a bone sec-tion is thick (Bachus and Bloebaum, 1992). Using a ran-dom grid system, five osteons were selected from eachimage. The public domain NIH image (v.1.61) software(http://rsb.info.nih.gov/nih-image/) was used to manuallytrace each of the five osteons. Similar to the processused in the ribs, these tracings were converted into digi-tized binary images that were used to measure most ofthe osteon parameters. As also described above for ribs,On.W.Th was calculated as the difference between themeasured outer (cement line) and HC radii (Fig. 7).

Osteon Size as a Possible Confounding Factor

To assess potential effects of osteon size on the rela-tionship between Ot.Lc.N/B.Ar and On.W.Th, Pearson’scorrelation coefficients were determined for these fourgroups in each bone type: (1) only smaller osteons with�150 lm diameter (in ribs also: �185 lm and �200 lm,see rationale below), (2) only smaller osteons � themean value of the data set, (3) only the larger osteons >the mean value of the data set (in ribs also: >185 lmand >200 lm), and (4) all osteons. Because of only 15osteons in the �150 lm group of ribs, the cutoff value

Fig. 6. Lateral-to-medial view of a left forelimb skeleton of an adulthorse showing the radius and third metacarpal (MC3). At near rightare the cross-sections showing image locations, and at far right arethe approximate typical ranges that the neutral axes (N.A.) moves dur-ing weight bearing. Several studies reporting in vivo strain data were

used to create these drawings (Turner et al., 1975; Rubin and Lanyon,1982; Schneider et al., 1982; Biewener et al., 1983a,b; Gross et al.,1992). C: Compression region (darkened); T: tension region (obliquelines); Cr: cranial; Cd: caudal; Dor: dorsal; Pal: palmar; Med: medial;and Lat: lateral.

Fig. 7. A: Tracing of an actual osteon showing the measured pa-rameters, which are listed below the osteon. B: Actual osteon con-verted to a perfect circle; below this drawing are the calculatedparameters. Ot.Lc.N are not shown in these drawings.

OSTEOCYTE DENSITY AND OSTEON WALL THICKNESS 1477

for this small/large osteon size distinction was raised to185 lm (similar to the largest mean diameter value ofthe limb bones) and also arbitrarily to 200 lm.

The correlation coefficients in each of the four sizegroups were also determined for (1) comparisons ofOt.Lc.N/B.Ar with On.W.Th/On.Dm (On.Dm ¼ osteonouter diameter), On.W.Th/On.Ar, HC.Ar/On.Ar, and B.Arand (2) comparisons among On.W.Th, On.Dm, On.Ar,and HC.Ar.

Using a one-way analysis of variance (ANOVA) design,paired comparisons of the Ot.Lc.N/B.Ar, (Ot.Lc.N/B.Ar)/On.W.Th, (Ot.Lc.N/B.Ar)/B.Ar, Haversian canal diameter(HC.Dm), On.Dm, and HC.Dm/On.Dm were evaluatedindependently within each bone type between the smalland large osteon groups. The parameters ‘‘(Ot.Lc.N/B.Ar)/On.W.Th’’ and ‘‘(Ot.Lc.N/B.Ar)/B.Ar’’ allowed forthe assessment of influences of the calculated linearmeasurement of wall thickness compared with the meas-ured wall ‘‘thickness’’ expressed as the area of the osteonwall (B.Ar; Fig. 7). In turn, this distinction allowed forthe detection of potential influences of osteon obliquity,which would be revealed by discrepancies between thesecalculated and measured parameters in the paired com-parison analyses (ANOVA and/or correlations). The crite-rion for statistical significance was P � 0.05 andvariability was expressed as the standard deviation.

RESULTS

Osteon diameters (On.Dm) varied greatly (Fig. 8), andthis was the dominant factor governing On.W.Thbecause the range of HC.Dm was much less (Fig. 9).Indeed, linear regression analysis showed that On.Dmaccounted for �89% of the variability in On.W.Th whenall osteons were considered in each bone type. In regres-sion analyses of small versus large osteons, On.Dmremained a dominant factor in governing On.W.Th butthe variability explained was somewhat less, rangingfrom 67% to 89%.

Correlation analyses of On.W.Th with On.Dm and ofOn.W.Th with On.Ar showed consistently strong rela-tionships that are expected between these parameters ineach of the seven bone types (data not shown). By con-

trast, positive correlations of HC.Ar with On.Ar weremore variable, with the ribs showing the strongest corre-lation (r ¼ 0.639 when all osteons were considered)when compared with the limb bones (range of r valuesin limb bones: 0.044–0.449). This reflects relativelygreater increases in HC.Ar with the increases in On.Arin the rib osteons.

Table 2 summarizes the results of correlationsbetween Ot.Lc.N/B.Ar and On.W.Th or On.W.Th/On.Dm.In general, these results revealed correlations that,although statistically significant (i.e., significantly differ-ent from no correlation), were typically weak—the vari-ability explained in these relationships ranged from�2% to 22%. As described below, these results neitherstrongly nor consistently support the hypotheses; the elkcalcanei are an obvious exception in terms of the smallversus large osteon analysis.

Figure 10 shows the diagrammatic representations of14 actual osteons that were selected from the equineradii and human ribs (seven osteons from each species).These represent the range of On.Dm from the smallestto the largest. Figure 11 shows the three representativeBSE images obtained from a deer calcaneus and horseradius.

Correlations of Ot.Lc.N/B.Ar With On.W.Th

In Table 2, the first row within each bone type liststhe results of correlation analyses of Ot.Lc.N/B.Ar withOn.W.Th using the data from all osteons as well as datasegregated by the osteon size. Each limb bone showed asignificant negative correlation between Ot.Lc.N/B.Arand On.W.Th (P < 0.001; r values typically �0.30 to�0.50). These results are also consistent with the magni-tude of the negative correlation shown previously in theosteons of sheep ulnae (r ¼ �0.39, P < 0.0001; Metzet al., 2003). Analysis of human rib data, however,showed a much weaker negative correlation (r ¼ �0.102,P ¼ 0.04).

Evaluation of Influences of Osteon Size

Results of correlation analyses in groups defined bythe small (�150 lm diameter; in ribs also: �185 and

Fig. 8. Histograms of frequency distributions of the outer diameters(lm) of all osteons analyzed in this study. The outer diameter isdefined by the cement line.

Fig. 9. Histograms of frequency distributions of the HC.Dm (lm) ofall osteons analyzed in this study.

1478 SKEDROS ET AL.

�200 lm), mean-small (�the mean On.Dm), and largeosteon (>the mean On.Dm; in ribs: >185 and >200 lm)distinctions are summarized in Table 2. These resultsgenerally show (elk calcanei are an exception) a possibleaffect of osteon size on the relationship of Ot.Lc.N/B.Arwith On.W.Th or with On.W.Th/On.Dm.

Results of these small versus large osteon analysesalso show that, in all cases, the absolute strengths of thecorrelations of Ot.Lc.N/B.Ar versus On.W.Th weregreater in the mean-small osteon group than in thelarger osteon group. However, as noted, this was not thecase in the elk calcanei when their small osteon groupwas defined by the �150 lm diameter cutoff: (1) �150lm: r ¼ �0.170, P ¼ 0.1 and (2) >164 lm: r ¼ �0.260, P< 0.01.

In ribs, an effect of On.Dm is suggested by resultsshowing that, when compared with the larger osteons,the smaller osteon groups (�185 and �200 lm) exhibitedincreases in the strength of the correlation of Ot.Lc.N/B.Ar with On.W.Th; for example: (1) �185 lm: r ¼�0.453, P < 0.001 and (2) >185 lm: r ¼ �0.066, P ¼ 0.2(Table 2).

Correlations supporting the second hypothesis (aneffect of On.Dm) in most bones examined are consistentwith the results of ANOVA paired comparisons. Thesecomparisons showed that (Ot.Lc.N/B.Ar)/On.W.Th and(Ot.Lc.N/B.Ar)/B.Ar were each greater in the smallerthan the larger osteon groups in all the seven bone types(P < 0.001; Table 3). In contrast to these findings, (1)the elk calcanei only showed the statistically significant

negative correlation of Ot.Lc.N/B.Ar with On.W.Th/On.Dm when all osteons are considered (i.e., the first hy-pothesis of the inverse relationship appears to be weaklysupported, but the second hypothesis of an osteon sizeeffect was rejected because these correlations were notsignificant in the small, mean-small, and large osteongroups; Table 2) and (2) the equine radii and equineMC3s showed a much weaker influence of osteon size,the correlation coefficients dealing with their largeosteon groups remained negative and significant buthave absolute magnitudes <0.1.

DISCUSSION

The results of this study do not provide clear or con-sistent support for an inverse relationship betweenosteocyte density and wall thickness of fully formed sec-ondary osteons. Although this conclusion is mainly basedon the correlations, which cannot directly elucidate spe-cific mechanisms or causation, we feel that a strong rela-tionship between osteon infilling and Ot.Lc.N/B.Ar, ifpresent, would have been detected in the five speciesthat were analyzed. However, this opinion reflects ourinitial thinking in terms of the hypothesis of Marottiand Martin that osteocytes repress infilling via signalingof a neural-like network. Alternatively, if sclerostinserves this repressive function, then it may be that theosteocytes closest to the forming central canal have thegreatest effect on the osteoblasts. In other words, non-uniform concentration of sclerostin across the osteonal

TABLE 2. Correlations of Ot.Lc.N/B.Ar with On.W.Th and On.W.Th/On.Dm in small, mean-small, mean-large, and all osteons from each speciesa

Species & bone On.Lc.N/B.Ar vs

Small osteonsMean-smallosteonsb

Mean-largeosteonsb All osteons

r P r P r P r P

Sheep calcaneus Small, �150 lm Small, �161 lm Large, >161 lm AllOn.W.Th �0.517 <0.001 �0.498 0.001 �0.195 0.1 �0.475c <0.001

On.W.Th/On.Dm �0.311 <0.01 �0.323 <0.01 �0.126 0.2 �0.402c <0.001Deer calcaneus Small, �150 lm Small, �152 lm Large, > 152 lm All

On.W.Th �0.238 <0.001 �0.230 <0.001 �0.216 <0.001 �0.337 <0.001On.W.Th/On.Dm �0.150 <0.01 �0.141 <0.01 0.060 0.3 �0.170c <0.001

Elk calcaneus Small, �150 lm Small, �164 lm Large, >164 lm AllOn.W.Th �0.170 0.1 �0.284 <0.001 �0.260 <0.01 �0.360 <0.001

On.W.Th/On.Dm �0.018 0.9 �0.090 0.3 �0.099 0.3 �0.163c 0.01Equine calcaneus Small, �150 lm Small, �181 lm Large, >181 lm All

On.W.Th �0.475 <0.001 �0.456 <0.001 �0.080 0.4 �0.467c <0.001On.W.Th/On.Dm �0.380 <0.01 �0.373 <0.001 �0.102 0.3 �0.361c <0.001

Equine radius Small, �150 lm Small, �186 lm Large, >186 lm AllOn.W.Th �0.384 <0.001 �0.403 <0.001 �0.130 <0.001 �0.386c <0.001

On.W.Th/On.Dm �0.303 <0.001 �0.310 <0.001 �0.091 <0.01 �0.301 <0.001Equine MC3 Small, <150 lm Small, �166 lm Large, >166 lm All

On.W.Th �0.442 <0.001 �0.430 <0.001 �0.152 <0.001 �0.419c <0.001On.W.Th/On.Dm �0.365 <0.001 �0.371 <0.001 �0.065 0.02 �0.340c <0.001

Human rib Small, �150 lm Small, �232 lm Large, >232 lm AllOn.W.Th �0.285 0.3 �0.167 0.02 �0.011 0.9 �0.102c 0.04

On.W.Th/On.Dm �0.196 0.5 �0.228 <0.01 0.026 0.7 �0.110c 0.03Small, �185 lm Small, �200 lm Large, >185 lm Large, >200 lm

On.W.Th �0.453 <0.001 �0.210 0.03 �0.066 0.2 �0.025 0.7On.W.Th/On.Dm �0.334 <0.01 �0.258 0.01 �0.043 0.4 �0.023 0.7

ar ¼ correlation coefficient.bThese small vs large distinctions are based on the mean diameter of the osteons in each bone type.cCases dealing with On.W.Th or On.W.Th/On.Dm where correlation coefficients differ more than the absolute value of 0.3and/or P values change from significant to nonsignificant (or vice versa) in the analyses of the six columns on the rightside of the table [mean-small (b), large (b), or all osteons]. See text for details.

OSTEOCYTE DENSITY AND OSTEON WALL THICKNESS 1479

wall (e.g., highest near the central canal) might be seenas a weak relationship in Ot.Lc.N/B.Ar versus On.W.Th/On.Dm, when linear correlations that are used to indi-rectly detect this effect are based on the osteocyte den-sity data from the entire osteon wall. The strengths ofthe correlations could also be influenced by an unrecog-nized effect of variations in osteocyte density from theouter margin of the osteon toward the central canal. Ina recent study using ultra-high-resolution micro-compu-terized tomography (Synchotron) to image human sec-ondary osteons (11 osteons; lateral aspect of proximalfemoral cortex; 25-year-old male), Hannah et al. (2010)show a clear trend of decreased numbers of osteocyticlacunae toward the central canal. If this trend is alsopresent in the osteons examined in our study, it couldplay a role in diminishing the strength of the correla-tions. Experimental data have also shown that convec-tive fluid flow can produce nonuniform concentrations ofproteins and other molecules that have the capacity tobe moved/transported through the interstitial micro-porosities of bone (Knothe Tate et al., 1998). In this per-spective, it seems plausible that the repressive effect ofa protein like sclerostin could be present, even thoughlittle or no correlation is detected between an osteon’sosteocyte density and wall thickness.

There are additional complexities in cellular dynamicsthat can influence the process of osteon infilling andthereby obscure ‘‘expected’’ correlations of osteocyte den-sity with wall thickness. For example, intercellular gap-junction communication is another way that osteocytes

can communicate with osteoblasts and potentially modu-late their responses to mechanical signals (Taylor et al.,2007). Osteoblast apoptosis may also be a factor that influ-ences the termination of bone formation during osteoninfilling (Parfitt, 1994; Qiu et al., 2010). Several studieshave additionally demonstrated the spatial relationshipbetween osteocyte apoptosis and osteoclastic bone resorp-tion (Aguirre et al., 2006; Hedgecock et al., 2007; Emertonet al., 2010), but the causal molecular or signaling linkhas not yet been identified between these two processes.The osteocyte has been thought to reduce osteoclasticbone resorption by secreting antiresorptive molecules (Guet al., 2005). The death of osteocytes causes the loss ofthese antiresorptive molecules, facilitating osteoclast pre-cursor recruitment, osteoclastogenesis, and bone resorp-tion (Gu et al., 2005). In addition, Kogianni et al. (2008)reported that apoptotic osteocytes could release signals,such as tumor necrosis factor-alpha (TNF-a), to initiateosteoclastogenesis and subsequent bone resorption.Osteoclasts, in turn, establish the size of the resorptionspace (i.e., the On.Dm), which cannot be larger than whatthe central capillary can supply to the tissue of the osteonwall (Ham, 1952). Clearly, much remains to be learnedabout how cellular and other biophysical factors interactin the regulation of osteon infilling.

Additional issues dealing with cell viability and age-related degradations in normal cell function point outwhat may be an important limitation of our study—wewere not able to quantify the percentage of viable cellsin the osteocytic lacunae. It could be that a greater

Fig. 10. Depicted are the diagrams based on the seven osteonsfrom the equine radii (at top) and seven osteons from the human ribs(at bottom). They range from the smallest to largest On.Dm. The inter-vening osteons are also the depictions of actual osteons spaced atapproximately equal intervals of diameters from this range. The

osteons are drawn to scale with contours that are smoothed to makethem circular; the distribution of osteocytes are shown to be quasirandom for illustrative purposes. The chart in the center lists the actualdata for each osteon (1, 2, 3, : : : , 7). Ot.Lc.N/B.Ar ¼ no./mm2 andother nonratio values are in micrometers.

1480 SKEDROS ET AL.

percentage of osteocytes were alive/functional in thesmall osteons compared with the large osteons, thusexplaining the differences between large and smallosteon responses. Unfortunately, it was not possible toascertain osteocyte viability in our slides. However, bycounting lacunae irrespective of their contents, ourintent was to understand the role of osteocytes in con-trolling osteoblasts during osteonal infilling; we wereinterested in osteocyte density as it was originally dur-ing the formation of the fully formed osteons that werequantified. Therefore, assuming that each osteocyte la-cuna had originally contained a living cell during osteonformation, we counted [as did Metz et al. (2003)] all

lacunae regardless of whether they contained evidenceof a cell body. Facts supporting the probability that ourbones had relatively smaller percentages of empty lacu-nae include (1) studies of osteocyte viability in youngerindividuals, such as those examined in most of the bonesin our study (the elk could be an exception, discussedbelow), show small percentages of empty lacunae (Qiuet al., 2006) and (2) we did not observe regions withlacunae that were plugged with hypermineralized tissuethat has been associated with aging, ischemia, or necro-sis (Frost, 1960; Currey, 1964; Kornblum and Kelly,1964; Jowsey, 1966; Parfitt, 1993). These observationsand the younger ages of the humans and most of theanimals used in our study suggest that the percentageof dead osteocytes is small. In the nonhuman bones, thisis also consistent with the observations of >90% lacunaoccupancy in the dorsal cortex of MC3s from thorough-bred and quarter horses with ages that are similar tothe horses used in the present study (Da Costa Gomezet al., 2005; Dr. Peter Muir, personal communication).Furthermore, the imaging methods used in our studyare sufficiently sensitive to detect the presence of micro-petrosis because of the obvious hypermineralized mate-rial that occludes the Ot.Lc.N in this condition (Busseet al., 2010). This phenomenon was not seen in any ofthe images, and it would also be unexpected because therelatively younger age of the humans and of a majorityof the animals that we studied.

Another possible limitation of the present study is thefact that osteon circularity was assumed in the stepsthat were taken to calculate the wall thickness from themeasured On.Ar and HC.Ar (Fig. 7). In a perfect circle,the diameter will have a very strong correlation withthe area [since area ¼ p(1=2Dm)2]. The difference betweencorrelations of Ot.Lc.N/B.Ar with On.W.Th/On.Dm whencompared with Ot.Lc.N/B.Ar with On.W.Th/On.Ar sug-gests that there is some obliquity to the osteons. Weattempted to minimize this variable [as did Metz et al.(2003)] by avoiding the analysis of obviously elliptical orotherwise unusual osteon variants. These exclusion cri-teria were deemed reasonable because these atypicalosteon morphologies represent a minority of the osteonsin the images used in the present study (Skedros et al.,2007b).

Given the small size of Ot.Lc.N, the sampling volumeof microscopic techniques could be an important con-founding factor in a two-dimensional analysis. Thismight seem to be an important consideration for thepresent study because two different microscopic techni-ques were used: epifluorescent microscopy was used toanalyze the human ribs and BSE microscopy was usedto analyze the nonhuman bones. However, the samplingvolumes of each technique do not extend beyond thebreadth of a typical osteocyte lacuna (Cane et al., 1982;Hannah et al., 2010). For the epifluorescent technique,this was assured by the use of blue-violet light becauseit penetrates only the first few microns of the bone sec-tion, allowing Ot.Lc.N to be quantified without projec-tion due to section thickness (Mori et al., 1997;Vashishth et al., 2000; Slyfield et al., 2009). A confound-ing effect of section thickness is also eliminated in BSEimaging because only a few microns of the bone surfaceare sampled when using the electron beam parametersused in this study (Bachus and Bloebaum, 1992; Skedroset al., 2005b).

Fig. 11. Three representative BSE images showing osteon size var-iations in (A) equine radius, caudal cortex at 100�; (B,C) deer calca-neus, plantar cortex at 50� (B) and 100� (C). The counting of Ot.Lc.Nin the 50� BSE images used in this study was facilitated with the aidof a magnifying lens.

OSTEOCYTE DENSITY AND OSTEON WALL THICKNESS 1481

TABLE

3.ANOVAsand

desc

rip

tivestatistics(m

eansand

standard

deviations)

Species&

bon

eOsteonsize

aOt.Lc.N/B.Ar

(no./m

m2)

(Ot.Lc.N/B.Ar)/

On.W

.Th

(Ot.Lc.N/B.Ar)/

B.Ar

HC.D

m(lm)

On.D

m(lm)

HC.D

m/

On.D

m

Sheepcalcaneus

Smallosteon

s(�

161lm,n¼

97)

756�

(265)

17,210�

(11,613)

74,487�

(70,378)

32�

(8)

131�

(21)

0.25�

(0.07)

Largeosteon

s(>

161lm,n¼

87)

599�

(154)

7,620�

(2,413)

21,896�

(8,112)

33�

(8)

195�

(23)

0.17�

(0.05)

Allosteon

s683�

(232)

12,675�

(9,825)

49,621�

(57,641)

32�

(8)

161�

(39)

0.21�

(0.07)

Deercalcaneus

Smallosteon

s(�

152lm,n¼

349)

695�

(253)

17,372�

(12,859)

75,898�

(87,217)

36�

(10)

125�

(20)

0.29�

(0.09)

Largeosteon

s(>

152lm,n¼

343)

575�

(172)

8,429�

(3,099)

24,961�

(10,493)

41�

(13)

182�

(26)

0.22�

(0.07)

Allosteon

s637�

(226)

13,074�

(10,505)

51,415�

(68,166)

38�

(12)

152�

(37)

0.26�

(0.09)

Elk

calcaneus

Smallosteon

s(�

164lm,n¼

136)

750�

(217)

15,906�

(7,143)

61,584�

(37,991)

37�

(10)

137�

(20)

0.27�

(0.08)

Largeosteon

s(>

164lm,n¼

108)

649�

(136)

8,856�

(2,657)

24,150�

(8,252)

44�

(15)

197�

(23)

0.23�

(0.04)

Allosteon

s705�

(192)

12,786�

(6,616)

45,015�

(34,336)

40�

(12)

163�

(37)

0.25�

(0.08)

Equin

ecalcaneus

Smallosteon

s(�

181lm,n¼

111)

692�

(192)

13,647�

(6,956)

51,478�

(37,694)

37�

(8)

147�

(23)

0.25�

(0.06)

Largeosteon

s(>

181lm,n¼

116)

553�

(126)

6,715�

(1,856)

16,931�

(5,560)

45�

(13)

214�

(27)

0.21�

(0.06)

Allosteon

s622�

(175)

10,104�

(6,112)

33,824�

(31,731)

41�

(12)

181�

(42)

0.23�

(0.06)

Equin

eradiu

sSmallosteon

s(�

186lm,n¼

1,200)

520�

(206)

11,118�

(8,358)

40,760�

(45,159)

45�

(11)

151�

(26)

0.30�

(0.08)

Largeosteon

s(>

186lm

,n¼

1,144)

419�

(128)

5,164�

(1,950)

12,146�

(5,154)

55�

(17)

224�

(31)

0.25�

(0.07)

Allosteon

s471�

(180)

8,212�

(6,817)

26,795�

(35,514)

50�

(15)

186�

(46)

0.28�

(0.08)

Equin

eMC3

Smallosteon

s(�

166lm,n¼

1,223)

644�

(297)

19,637�

(25,810)

76,257�

(107,414)

47�

(12)

132�

(25)

0.36�

(0.01)

Largeosteon

s(>

166lm,n¼

1,235)

496�

(148)

7,101�

(2,841)

18,460�

(8,019)

53�

(16)

199�

(27)

0.27�

(0.08)

Allosteon

s570�

(246)

13,338�

(19,357)

47,217�

(81,278)

50�

(11)

166�

(43)

0.32�

(0.11)

Human

rib

Smallosteon

s(�

200lm,n¼

111)

865�

(138)

13,087�

(3,376)

40,744�

(14,166)

37�

(10)

173�

(19)

0.22�

(0.06)

Largeosteon

s(>

200lm,n¼

287)

842�

(125)

8,493�

(1,780)

18,125�

(5,468)

53�

(14)

256�

(34)

0.21�

(0.05)

Smallosteon

s(�

232lm,n¼

191)

856�

(135)

11,741�

(3,230)

33,793�

(13,831)

40�

(10)

191�

(27)

0.21�

(0.05)

Largeosteon

s(>

232lm,n¼

207)

841�

(123)

7,960�

(1,474)

15,796�

(3,853)

57�

(14)

271�

(28)

0.21�

(0.05)

Allosteon

s848�

(129)

9,775�

(3,114)

24,433�

(13,428)

49�

(15)

232�

(48)

0.21�

(0.05)

¼Grayed

cellsindicate

statisticallysignificantdifferences(P

�0.05)betweenlargeandsm

allosteon

sin

thesa

mecolumnandwithin

thebon

etype.

aSmallosteon

sin

this

columnare

basedon

themea

ndiameter

inea

chbon

etype;

a200lm

cutoff

isalsosh

ownfortheribs.

Largeosteon

sare

defi

ned

asgreater

thanthemea

ndiameter

oftheosteon

sin

each

bon

etype.

1482 SKEDROS ET AL.

Clearly, several factors influence osteon morphology,including one of the dominant variables demonstrated inthis study—On.Dm. The influence of On.Dm and therather weak correlations shown herein could be consid-ered in terms of Martin’s (2000a) caution that an osteo-cyte-network inhibitory signal at the HC surface wouldbe very sensitive to the efficiency with which it can bepassed from one osteocyte to another (Fig. 3). Spatialsignal decay/plateau may help to explain our resultsshowing that correlations of Ot.Lc.N/B.Ar with On.W.Th/On.Dm were significant in the smaller osteons (in allbut elk calcanei; discussed further below) but becamenonsignificant in osteons with diameters greater than�150–190 lm. Alternatively, as mentioned above, thesedata could be explained by spatial variations in the effi-ciency of a signaling protein that is ‘‘transported’’through the microporosities of bone matrix and/or byother unrecognized complexities of cellular regulation ofinfilling.

In addition to the ‘‘transport’’ of sclerostin towardosteoblasts, the dynamics of nutrient delivery duringosteon infilling must be considered in the perspective ofthe differences shown in the present study betweensmaller and larger osteons. For example, if during theearlier phase of infilling, the outermost osteocytes ofsmaller osteons more quickly become nutrient deficientwhen compared with the outermost portions of largerosteons [i.e., greater HC surface area for nutrientexchange in larger osteons], then the relative degree of‘‘starvation’’ of the outermost osteocytes may be rela-tively greater in smaller osteons during their initialinfilling. If this occurs, then the inhibitory signal mightinfluence the duration of infilling and/or the formationperiod differently between these groups. If correct, thismight help explain our data showing that the HC.Dm/On.Dm ratio is greatest in the smaller osteon group (P <0.05) in all of the limb bones. This suggests that thehypothesized inhibitory signal is relatively deficient and/or becomes less precise in larger osteons. To test this hy-pothesis, additional studies of the dynamics of osteonformation are needed to evaluate potential differences inappositional rates, osteocyte densities, temporal/spatialvariations in sclerostin concentration, and formationperiods between early, middle, and late infilling phasesof osteons from a broad size range.

In addition to considering the possibility that osteocyterepression influences the extent of osteon infilling, an in-dependent effect of infilling rate on osteocyte densityshould be considered. On.Dm might also be important inthis context, because relatively faster rates of osteoninfilling have been shown in larger versus smallerosteons in dog ribs (r ¼ 0.52; Ilnicki et al., 1966), doglimb bones (e.g., femur; r ¼ 0.87; Lee, 1964), rabbit limbbones (Jee and Arnold, 1954), and human limb and ribbones (Polig and Jee, 1990). In turn, data from sheepulnae reported by Metz et al. (2003) support the possibil-ity that infilling rate, expressed as mineral appositionalrate (MAR), is positively correlated with osteocyte den-sity. Though statistically significant, this effect wasweak as shown by the correlation between Ot.Lc.N/B.Arand normalized intracortical MAR (r2 ¼ 0.12, P ¼0.0001). Taken together, these data might suggest thatthe density of osteocytes remaining after osteonal infill-ing is, at best, weakly influenced by the positive rela-tionship between MAR and osteon size. However, a

recent rigorous study of the relationships of modelingand remodeling with osteocyte density and apoptosis inrabbit tibia mid-shafts did not find significant relation-ships between osteocyte lacuna or osteocyte densitiesand intracortical (osteonal) or surface bone formationrates (Hedgecock et al., 2007). In this perspective, itseems unlikely that the osteocyte density differencesshown in the present study are strongly influenced bythe relationship between MAR and On.Dm.

Results in elk calcanei showing no correlations insmall or large osteons (Ot.Lc.N/B.Ar vs. On.W.Th orOn.W.Th/On.Dm) are anomalous in view of the correla-tions shown consistently (although weak) among theseparameters in the smaller osteons in the other six bonetypes. This exception could reflect a flaw in the hypothe-sis that osteon infilling is regulated, to some extent, byosteocyte density. This possibility can only be resolvedby the additional experimental studies that control forfactors that are not apparent in the present study. Forexample, in our sample of elk calcanei, it is possible thatthere is a significant influence of aging on osteon infill-ing dynamics, which has been described in the other spe-cies (e.g., decreasing osteon size with age; Lee, 1964;Ilnicki et al., 1966; Pirok et al., 1966; Britz et al., 2009).This possibility is consistent with our elk being larger/older bulls, reflecting the fact that elk hunters are typi-cally interested in shooting trophy-sized animals. Conse-quently, among all of the animals studied, the elk mighthave been, in general, relatively older in terms of theirtypical life expectancy. Additionally, data in adulthuman femora show that osteon diameter is inverselyrelated to body weight (Britz et al., 2009). These datasupport the idea that osteons (and remodeling processes)are dictated (to some degree) by load bearing or ‘‘usage.’’Perhaps the apparently anomalous results in the elkbones could be due to the fact that these are quite largeanimals, older, and very active compared with the otherbones.

The fact that the ribs showed the weakest correlationsof Ot.Lc.N/B.Ar with On.W.Th or On.W.Th/On.Dm (allosteons considered) might reflect important species dif-ferences in osteocyte/osteon physiology and morphologyor could reflect the fundamental differences in these con-texts between ribs and limb bones. There are data sup-porting the possibility of significant differences inmechanosensitivity between osteocytes of the axial (e.g.,ribs and calvarium) and osteocytes of the appendicular(limb) skeleton (Raab et al., 1991; Tommerup et al.,1993; Rawlinson et al., 1995; Vatsa et al., 2008). In thisperspective, the thresholds for metabolic activity of bonecells in ribs are probably fundamentally different whencompared with the appendicular bones—ribs are moresensitive to hormonal changes associated with lactationand other highly mineral-dependent demands, whileappendicular bones are less so (Vajda et al., 1999). Infact, expression of sclerostin is regulated by both me-chanical stimulation and PTH, suggesting that both me-chanical and metabolic influences may signal throughosteocytes to inhibit bone formation by osteoblasts onthe surface of the bone (Bellido et al., 2005; Roblinget al., 2008; Sims and Gooi, 2008). Consequently, insome species or circumstances, metabolic demands couldbe sufficient to affect osteonal morphology and biome-chanics differently in ribs when compared with the limbbones. These considerations might prove helpful in

OSTEOCYTE DENSITY AND OSTEON WALL THICKNESS 1483

explaining why in this study the osteons of the ribs,compared with the osteons of limb bones, are generallyrelatively larger, have higher numbers of Ot.Lc.N, havesimilar HC.Dm/On.Dm ratios in small and large osteons,and exhibit relatively lower correlations of Ot.Lc.N/B.Arwith On.W.Th/On.Dm (Table 2 and Fig. 10). Althoughthe results of computational analyses and some histo-morphometric studies of various limb bones showinverse relationships between mechanical stress andOn.Dm (i.e., higher stress produces smaller osteons; vanOers et al., 2008; Britz et al., 2009), anthropologicalstudies of osteon cross-sectional dimensions from rib andfemora of the same individuals show that rib osteons arenot consistently larger (Pfeiffer et al., 2006).

In summary, the results of this study do not provideclear or consistent support for the hypothesized inverserelationship between osteocyte density and wall thick-ness of fully formed secondary osteons. However, linearcorrelation analyses could fail to detect this evidence,although indirect, of osteocyte-based repression of osteoninfilling if the signal is spatially nonuniform (e.g.,increased near the central canal).

ACKNOWLEDGEMENTS

The authors thank Wm. Erick Anderson and JaxonHoopes for assisting with technical aspects of this study,Roy D. Bloebaum for laboratory support, and Dr. KenHunt for his criticisms of the manuscript.

LITERATURE CITED

Aguirre JI, Plotkin LI, Stewart SA, Weinstein RS, Parfitt AM, Man-olagas SC, Bellido T. 2006. Osteocyte apoptosis is induced byweightlessness in mice and precedes osteoclast recruitment andbone loss. J Bone Miner Res 21:605–615.

Bachus KN, Bloebaum RD. 1992. Projection effect errors in bioma-terials and bone research. Cells Mater 2:347–355.

Bellido T, Ali AA, Gubrij I, Plotkin LI, Fu Q, O’Brien CA, Manola-gas SC, Jilka RL. 2005. Chronic elevation of parathyroid hormonein mice reduces expression of sclerostin by osteocytes: a novelmechanism for hormonal control of osteoblastogenesis. Endocri-nology 146:4577–4583.

Biewener AA, Thomason J, Goodship A, Lanyon LE. 1983a. Bonestress in the horse forelimb during locomotion at different gaits: acomparison of two experimental methods. J Biomech 16:565–576.

Biewener AA, Thomason J, Lanyon LE. 1983b. Mechanics of locomo-tion and jumping in the forelimb of the horse (Equus): in vivo stressdeveloped in the radius and metacarpus. J Zool Lond 201:67–82.

Britz HM, Thomas CD, Clement JG, Cooper DM. 2009. The relationof femoral osteon geometry to age, sex, height and weight. Bone45:77–83.

Burger EH, Klein-Nulend J. 1999. Mechanotransduction in bone—role of the lacunocanalicular network. FASEB J 13:S101–S112.

Busse B, Djonic D, Milovanovic P, Hahn M, Puschel K, Ritchie RO,Djuric M, Amling M. 2010. Decrease in the osteocyte lacunar den-sity accompanied by hypermineralized lacunar occlusion revealsfailure and delay of remodeling in aged human bone. Aging Cell9:1065–1075.

Cane V, Marotti G, Volpi G, Zaffe D, Palazzini S, Remaggi F, MugliaMA. 1982. Size and density of osteocyte lacunae in differentregions of long bones. Calcif Tissue Int 34:558–563.

Cooper D, Thomason C, Clement J, Hallgrimsson B. 2006. Three-dimensional microcomputed tomography imaging of basic multi-cellular unit-related resorption spaces in human cortical bone.Anat Rec A: Discov Mol Cell Evol Biol 288:806–816.

Currey JD. 1964. Some effects of aging in human Haversian sys-tems. J Anat Lond 98:69–75.

Da Costa Gomez TM, Barrett JG, Sample SJ, Radtke CL, KalscheurVL, Lu Y, Markel MD, Santschi EM, Scollay MC, Muir P. 2005.Up-regulation of site-specific remodeling without accumulation ofmicrocracking and loss of osteocytes. Bone 37:16–24.

Emerton KB, Hu B, Woo AA, Sinofsky A, Hernandez C, MajeskaRJ, Jepsen KJ, Schaffler MB. 2010. Osteocyte apoptosis and con-trol of bone resorption following ovariectomy in mice. Bone46:577–583.

Emmanual J, Hornbeck C, Bloebaum RD. 1987. A polymethyl meth-acrylate method for large specimens of mineralized bone withimplants. Stain Technol 62:401–410.

Frost HM. 1960. In vivo osteocyte death. Am J Orthop 42-A:138–143.

Frost HM. 1990. Skeletal structural adaptations to mechanicalusage (SATMU). 2. Redefining Wolff ’s law: the remodeling prob-lem. Anat Rec 226:414–422.

Frost HM. 2001. From Wolff ’s law to the Utah paradigm: insightsabout bone physiology and its clinical applications. Anat Rec262:398–419.

Gross TS, McLeod KJ, Rubin CT. 1992. Characterizing bone straindistribution in vivo using three triple rosette strain gauges. J Bio-mech 25:1081–1087.

Gu G, Mulari M, Peng Z, Hentunen TA, Vaananen HK. 2005. Deathof osteocytes turns off the inhibition of osteoclasts and triggerslocal bone resorption. Biochem Biophys Res Commun 335:1095–1101.

Ham AW. 1952. Some histophysiological problems peculiar to calci-fied tissues. J Bone Joint Surg Am 24-A-3:701–728.

Hannah KM, Thomas CD, Clement JG, De Carlo F, Peele AG. 2010.Bimodal distribution of osteocyte lacunar size in the human femo-ral cortex as revealed by micro-CT. Bone 47:866–871.

Hedgecock NL, Hadi T, Chen AA, Curtiss SB, Martin RB, Hazel-wood SJ. 2007. Quantitative regional associations between remod-eling, modeling, and osteocyte apoptosis and density in rabbittibial midshafts. Bone 40:627–637.

Herman BC, Cardoso L, Majeska RJ, Jepsen KJ, Schaffler MB. 2010.Activation of bone remodeling after fatigue: differential response tolinear microcracks and diffuse damage. Bone 47:766–772.

Ilnicki L, Epker BN, Frost HM, Hattner R. 1966. The radial rate ofosteon closure evaluated by means of in vivo tetracycline labelingin beagle dog rib. J Lab Clin Med 67:447–454.

Irie K, Ejiri S, Sakakura Y, Shibui T, Yajima T. 2008. Matrix miner-alization as a trigger for osteocyte maturation. J Histochem Cyto-chem 56:561–567.

Jee WS, Arnold JS. 1954. Rate of individual Haversian system for-mation. Anat Rec 118:315.

Jowsey J. 1966. Studies of Haversian systems in man and some ani-mals. J Anat 100:857–864.

Knothe Tate ML, Niederer P, Knothe U. 1998. In vivo tracer trans-port through the lacunocanalicular system of rat bone in an envi-ronment devoid of mechanical loading. Bone 22:107–117.

Kogianni G, Mann V, Noble BS. 2008. Apoptotic bodies convey activ-ity capable of initiating osteoclastogenesis and localized bonedestruction. J Bone Miner Res 23:915–927.

Kornblum SS, Kelly PJ. 1964. The lacunae and Haversian canals intibial cortical bone from ischemic and non-ischemic limbs. A com-parative microradiographic study. J Bone Joint Surg Am 46:797–810.

Lanyon LE. 1974. Experimental support for the trajectorial theoryof bone structure. J Bone Joint Surg 56B:160–166.

Lee WR. 1964. Appositional bone formation in canine bone: a quan-titative microscopic study using tetracycline markers. J Anat98:665–677.

Marotti G. 1996. The structure of bone tissues and the cellular con-trol of their deposition. Ital J Anat Embryol 101:25–79.

Marotti G, Ferretti M, Muglia MA, Palumbo C, Palazzini S. 1992. Aquantitative evaluation of osteoblast–osteocyte relationships ongrowing endosteal surface of rabbit tibiae. Bone 13:363–368.

Martin RB. 2000a. Does osteocyte formation cause the nonlinearrefilling of osteons. Bone 26:71–78.

Martin RB. 2000b. Toward a unifying theory of bone remodeling.Bone 26:1–6.

1484 SKEDROS ET AL.

Mason MW, Skedros JG, Bloebaum RD. 1995. Evidence of strain-mode-related cortical adaptation in the diaphysis of the horse ra-dius. Bone 17:229–237.

Metz LN, Martin RB, Turner AS. 2003. Histomorphometric analysisof the effects of osteocyte density on osteonal morphology andremodeling. Bone 33:753–759.

Mori S, Harruff R, Ambrosius W, Burr DB. 1997. Trabecular bone vol-ume and microdamage accumulation in the femoral heads ofwomen with and without femoral neck fractures. Bone 21:521–526.

Parfitt AM. 1983. The physiologic and clinical significance of bonehistomorphometric data. In: Recker RR, editor. Bone histomorph-ometry: techniques and interpretation. Boca Raton: CRC Press,Inc. p 143–224.

Parfitt AM. 1993. Bone age, mineral density, and fatigue damage.Calcif Tissue Int 53 (Suppl 1):S82–S85; discussion S85–S86.

Parfitt AM. 1994. Osteonal and hemi-osteonal remodeling: the spa-tial and temporal framework for signal traffic in adult humanbone. J Cell Biochem 55:273–286.

Pfeiffer S, Crowder C, Harrington L, Brown M. 2006. Secondaryosteons and Haversian canal dimensions as behavioral indicators.Am J Phys Anthropol 131:460–468.

Pirok DJ, Ramser JR, Takahashi H, Villanueva AR, Frost HM.1966. Normal histological, tetracycline and dynamic parametersin human, mineralized bone sections. Henry Ford Hosp Med J14:195–218.

Polig E, Jee WS. 1990. A model of osteon closure in cortical bone.Calcif Tissue Int 47:261–269.

Power J, Loveridge N, Rushton N, Parker M, Reeve J. 2002. Osteo-cyte density in aging subjects is enhanced in bone adjacent toremodeling Haversian systems. Bone 30:859–865.

Power J, Noble BS, Loveridge N, Bell KL, Rushton N, Reeve J.2001. Osteocyte lacunar occupancy in the femoral neck cortex: anassociation with cortical remodeling in hip fracture cases and con-trols. Calcif Tissue Int 69:13–19.

Qiu S, Fyhrie DP, Palnitkar S, Rao DS. 2003. Histomorphometricassessment of Haversian canal and osteocyte lacunae in different-sized osteons in human rib. Anat Rec 272A:520–525.

Qiu S, Rao DS, Palnitkar S, Parfitt AM. 2006. Differences in osteo-cyte and lacunar density between Black and White Americanwomen. Bone 38:130–135.

Qiu S, Rao DS, Palnitkar S, Parfitt AM. 2010. Dependence of boneyield (volume of bone formed per unit of cement surface area) onresorption cavity size during osteonal remodeling in human rib:implications for osteoblast function and the pathogenesis of age-related bone loss. J Bone Miner Res 25:423–430.

Raab DM, Crenshaw TD, Kimmel DB, Smith EL. 1991. A histomor-phometric study of cortical bone activity during increased weight-bearing exercise. J Bone Miner Res 6:741–749.

Rawlinson SD, Mosley JR, Suswillo RF, Pitsillides AA, Lanyon LE.1995. Calvarial and limb bone cells in organ and monolayer cul-ture do not show the same early response to dynamic mechanicalstrains. J Bone Miner Res 10:1225–1232.

Riddle RC, Donahue HJ. 2009. From streaming-potentials to shearstress: 25 years of bone cell mechanotransduction. J Orthop Res27:143–149.

Robling AG, Niziolek PJ, Baldridge LA, Condon KW, Allen MR,Alam I, Mantila SM, Gluhak-Heinrich J, Bellido TM, Harris SE,Turner CH. 2008. Mechanical stimulation of bone in vivo reducesosteocyte expression of Sost/sclerostin. J Biol Chem 283:5866–5875.

Rubin CT, Lanyon LE. 1982. Limb mechanics as a function of speedand gait: a study of functional strains in the radius and tibia ofhorse and dog. J Exp Biol 101:187–211.

Schneider RK, Milne DW, Gabel AA, Groom JJ, Bramlage LR. 1982.Multidirectional in vivo strain analysis of the equine radius andtibia during dynamic loading with and without a cast. Am J VetRes 43:1541–1550.

Sims NA, Gooi JH. 2008. Bone remodeling: multiple cellular inter-actions required for the coupling of bone formation and resorp-tion. Semin Cell Dev Biol 19:444–451.

Skedros JG. 2005. Osteocyte lacuna population densities in sheep,elk and horse calcanei. Cells Tissues Organs 181:23–37.

Skedros JG, Bloebaum RD, Bachus KN, Boyce TM, Constantz B.1993. Influence of mineral content and composition of gray levelsin backscattered electron images of bone. J Biomed Mater Res27:57–64.

Skedros JG, Bloebaum RD, Mason MW, Bramble DM. 1994a. Analy-sis of a tension/compression skeletal system: possible strain-spe-cific differences in the hierarchical organization of bone. Anat Rec239:396–404.

Skedros JG, Grunander TR, Hamrick MW. 2005a. Spatial distribu-tion of osteocyte lacunae in equine radii and third metacarpals:considerations for cellular communication, microdamage detectionand metabolism. Cells Tissues Organs 180:215–236.

Skedros JG, Holmes JL, Vajda EG, Bloebaum RD. 2005b. Cementlines of secondary osteons in human bone are not mineral-defi-cient: new data in a historical perspective. Anat Rec A: DiscovMol Cell Evol Biol 286:781–803.

Skedros JG, Mason MW, Bloebaum RD. 1994b. Differences in osteo-nal micromorphology between tensile and compressive cortices ofa bending skeletal system: indications of potential strain-specificdifferences in bone microstructure. Anat Rec 239:405–413.

Skedros JG, Mason MW, Nelson MC, Bloebaum RD. 1996. Evidenceof structural and material adaptation to specific strain features incortical bone. Anat Rec 246:47–63.

Skedros JG, Sorenson SM, Hunt KJ, Holyoak JD. 2007a. Ontoge-netic structural and material variations in ovine calcanei: a modelfor interpreting bone adaptation. Anat Rec 290:284–300.

Skedros JG, Sorenson SM, Jenson NH. 2007b. Are distributions ofsecondary osteon variants useful for interpreting load history inmammalian bones? Cells Tissues Organs 185:285–307.

Skedros JG, Su SC, Bloebaum RD. 1997. Biomechanical implica-tions of mineral content and microstructural variations in corticalbone of horse, elk, and sheep calcanei. Anat Rec 249:297–316.

Slyfield CR, Jr., Niemeyer KE, Tkachenko EV, Tomlinson RE,Steyer GG, Patthanacharoenphon CG, Kazakia GJ, Wilson DL,Hernandez CJ. 2009. Three-dimensional surface texture visualiza-tion of bone tissue through epifluorescence-based serial block faceimaging. J Microsc 236:52–59.

Su SC, Skedros JG, Bachus KN, Bloebaum RD. 1999. Loading con-ditions and cortical bone construction of an artiodactyl calcaneus.J Exp Biol 202 (Pt 22):3239–3254.

Taylor AF, Saunders MM, Shingle DL, Cimbala JM, Zhou Z, Dona-hue HJ. 2007. Mechanically stimulated osteocytes regulate osteo-blastic activity via gap junctions. Am J Physiol Cell Physiol292:C545–C552.

Tommerup LJ, Raab DM, Crenshaw TD, Smith EL. 1993. Doesweight-bearing exercise affect non-weight-bearing bone? J BoneMiner Res 8:1053–1058.

Turner AS, Mills EJ, Gabel AA. 1975. In vivo measurement of bonestrain in the horse. Am J Vet Res 36:1573–1579.

Turner CH, Warden SJ, Bellido T, Plotkin LI, Kumar N, Jasiuk I,Danzig J, Robling AG. 2009. Mechanobiology of the skeleton. SciSignal 2:pt3.

Vajda EG, Kneissel M, Muggenburg B, Miller SC. 1999. Increasedintracortical bone remodeling during lactation in beagle dogs. BiolReprod 61:1439–1444.

Van Bezooijen RL, Roelen BA, Visser A, van der Wee-Pals L, deWilt E, Karperien M, Hamersma H, Papapoulos SE, ten Dijke P,Lowik C. 2004. Sclerostin is an osteocyte-expressed negative regu-lator of bone formation, but not a classical bmp antagonist. J ExpMed 199:805–814.

van Oers RF, Ruimerman R, Tanck E, Hilbers PA, Huiskes R. 2008.A unified theory for osteonal and hemi-osteonal remodeling. Bone42:250–259.

Vashishth D, Verborgt O, Divine G, Schaffler MB, Fyhrie DP. 2000.Decline in osteocyte lacunar density in human cortical bone is associ-ated with accumulation of microcracks with age. Bone 26:375–380.

Vatsa A, Breuls RG, Semeins CM, Salmon PL, Smit TH, Klein-Nulend J. 2008. Osteocyte morphology in fibula and calvaria—isthere a role for mechanosensing? Bone 43:452–458.

Weinbaum S, Guo P, You L. 2001. A new view of mechanotransduc-tion and strain amplification in cells with microvilli and cell proc-esses. Biorheology 38:119–142.

OSTEOCYTE DENSITY AND OSTEON WALL THICKNESS 1485